Konrad Heidler

I found the problem... When correctly specifying `visibleRect: [0, 1000, 1000, 0]`, the behavior is as intended. Should there be some graceful handling of "negative height" for people who intuitively...

Thanks for getting back to me :) Yeah, I get that my proposal both breaks backward compatibility and introduces possible confusion for flipping axes... It was just an idea of...

Wow this issue has been around for half a year? :sweat_smile: I just stumbled across the same issue, the following behaviour is super weird: ```python a = jnp.zeros([1, 8, 8,...

Welp, I guess the docs **do** indeed specify that `channel_axis` is only used if `window_shape`/`strides` are an integer. It's still super confusing, but actually documented.

Currently, ANSI escape codes will already work, e.g. ```julia iter = ProgressBar(1:100) for i in iter # ... set_multiline_postfix(iter, "\u001b[31mred: \u001b[32mgreen\u001b[37m") end ``` However, this is not a great API...

Okay that's convenient, I guess we won't need a new API from the ProgressBars.jl side for that then. The ANSI escapes will currently mess up the character-counting for the multiline_postfix:...

Just released v1.5.0 that fixes colorful multiline postfixes

In newer versions of earthengine-api, this actually leads to an error message: `ee.Initialize: no project found. Call with project= or see http://goo.gle/ee-auth` This even persists when the user correctly initializes...

1. They are in fact two independent loss terms. The `auto_weight_bce` loss function calculates both in parallel for efficiency. 2. Beta is equivalent to the term $|Y_+| / |Y_+ \cup...

The HSV-code is to blame here... Maybe jitter in a R^2 (polar) representation of HS-Space?